Transactions on the Built Environment vol 66, © 2003 WIT Press, www.witpress.com, ISSN 1743-3509 Finite element modeling of Guastavino tiled arches E. P. Saliklis, S. J. Kurtz & S. V.Furnbach Dept. of Civil and Environmental. Engineering, Lafayette College, USA Abstract An investigation of Rafael Guastavino's arches has been conducted by means of finite element modeling and laboratory experimentation. A novel method of modeling laminated masonry tile construction via the finite element method has been devised. This technique takes advantage of the layered shell element features found in commercially available finite element programs. Historical Guastavino tiles have been tested to obtain material properties. These modem techniques have been employed in conjunction with Guastavino's original empirical design criteria to provide a better understanding of these historically significant structures. 1 Introduction Rafael Guastavino, born in 1842, emigrated to the United States to establish the Guastavino Fireproof Construction Company. The fascinating architectural legacy of Guastavino and his son (also Rafael Guastavino) has received scholarly attention [1],[2] but the mechanics of his designs have not garnered similar attention from structural engineers. The thin laminated tile construction that the elder Guastavino used in hundreds of structures in the Eastern United States had its roots in his native Catalan's indigenous vaulting badition. Before emigrating to the United States, Rafael Guastavino designed such laminated vaulting in Barcelona [3]. While it has been suggested that Guastavino came to the United States to utilize superior cements in his mortars [4], others claim it is more feasible to say that he emigrated because of his faith in the American construction industry and its ability to produce consistent and high quality materials [5]. Transactions on the Built Environment vol 66, © 2003 WIT Press, www.witpress.com, ISSN 1743-3509 258 Strrcctrrral Srudir~,Rrpam atrd Marnrorancr of Hrriragr Aldzirrcruw V111 Previous to ths study, the engineering material properties of Guastavino tiles have not been quantitatively analyzed, although Lane has performed chemical analyses of the mortars [5]. Much mystery surrounds the formulation of the mortar since many Guastavino vaults were erected without centering or scaffolding. Lane [5] found that many mortars were traditional, simple mixtures of Portland cement and sand, and Parks and Neumann report one part Portland cement to two and one half parts sharp clean sand (specifically Cow Bay sand, a sharp angular sand quarried in Long Island New York [2]. In fact, although Guastavino filed a number of U.S. patents for his fireproof "cohesive" laminated tile construction, none of the patents describe the properties of the mortar, which gives firther credence to the theory that the mortar was completely traditional. Even more mystery surrounds the design methods that Guastavino used in his vaults, arches, domes and stairs. Guastavino claimed that his arches and vaults produced little lateral thrust under gravity loading, and he produced very few design formula in his largely promotional book "Cohesive Construction" [6]. Throughout the history of the Guastavino Company, there appears to have been only one engineer responsible for all design calculations, and his records were lost after the closing of the company in 1962. During the early 190OYs,structural engineers were just beginning to understand the mechanics of thm shell structures, and theories were available only for simple geometries such as hemispheres in the 1930's. Guastavino combined intuition with empiricism to design spectacular spaces of extreme thinness, doing so with common materials. Th~spaper will analyze some typical designs by means of the finite element method. 2 Mechanical testing of tiles 2.1 Description of recovered tiles A total of fifteen Guastavino tile fragments were obtained, ranging in size from about 5160 mm2,up to 29030 mm2,each with a nominal thickness of 2.54 mm. As shown in Figure 1, the tiles were not all the same. In particular, there appeared to be three different groove patterns manufactured in the tiles' surfaces, varying from small (165 grooves per meter) to medium (138 grooves per meter) to large (106 grooves per meter) grooves. Hence, the first objective of the mechanical testing program was to determine if there were significant differences in mechanical properties, from tile to tile. Transactions on the Built Environment vol 66, © 2003 WIT Press, www.witpress.com, ISSN 1743-3509 Figure 1: Recovered Tile Specimens. 2.2 Samples and testing A total of five tiles were tested. Due to the small amount of available tile specimens, it was decided that greatest use would be made from non-destructive testing, which generally requires smaller specimen sizes. Each of the tested tiles was ground smooth using a diamond abrasive wheel. Then, a water-cooled tile saw was used to cut mechanical testing samples from each tile fragment. Samples were cut in two perpendicular directions, for the purpose of evaluating the anisotropic elastic properties of the tiles. As indicated in Figure 2, the direction of the surface grooves is designated as the X-direction, the direction perpendicular to the grooves is designated as the Y-drection, and the direction through the tile thickness is designated as the Z-direction. Figure 2: Tile Coordinate System. Transactions on the Built Environment vol 66, © 2003 WIT Press, www.witpress.com, ISSN 1743-3509 260 Strrcctrrral Srudir~,Rrpam atrd Marnrorancr of Hrriragr Aldzirrcruw V111 2.3 Methods Mechanical testing consisted of; longitudinal and transverse dynamic elastic modulus testing, monotonic compression testing to determine the ultimate cornpressive strength, and monotonic flexure testing to determine the ultimate tensile strength. The longitudinal and transverse dynamic elastic modulus tests were conducted in accordance with ASTM C215 [7], which are free vibration tests, measured via a bonded accelerometer. Data was acquired by a 100 kHz PC- based card, with spectrum analysis provided within the Lab View [g] software package. Typically, the fundamental frequencies of vibration were determined three ways for each sample: the longitudinal (axial) mode, and the transverse (bending) modes, evaluated about both principal bending axes. In every case, the elastic moduli, evaluated three different ways, were in close agreement with one another. In the monotonic compression tests, a manually-controlled hydraulic testing machine was used to record the ultimate strength of tile samples. For the monotonic flexure testing a manual displacement-controlled testing machine was used to record the ultimate flexural strength of tile samples, in 3-point bending, over spans of 51 mm or 102 mm, where the larger span was used for thicker samples (typically equal to the ground tile thickness of about 20 mm), while the smaller span was used for thinner specimens (typically, 13 mm). 2.4 Mechanical testing results 2.4.1 Dynamic tests for Young's modulus A total of 38 free vibration tests were conducted on tile samples in the X- direction, while 22 tests were conducted in the Y-direction. For individual tiles, statistical variation was small, with the coefficient of variation for tests within a tile specimen usually less than 10%. This variation tended to be smaller fox testing in the X-direction, than in the Y-direction. The most notable result was consistent orthotropic properties. On average, the dynamic elastic modulus E, was 21520 MPa in the X-direction, but E, was only 12000 MPa in the Y-direction. For all tiles, E, exceeded E, by a factor of about 1.8. Examining the statistical variation in both directions, it is clear that the orthotropic properties are significant throughout all of the tested tiles. However, despite the presence of orthotropic properties, it is necessary to use an average value for E of about 16548 MPa in structural models, because it is assumed that the tiles were randomly oriented in the structure. At the onset of testing, it was hypothesized that the hfferences in surface appearance, namely the sizes of surface grooves, may correlate with differences in properties. Regarding ths hypothesis, Table 1 indicates that the tile-to-tile variations are not much larger than the variation of test results within a tile. This suggests that surface appearance does not correlate with X-direction properties. Transactions on the Built Environment vol 66, © 2003 WIT Press, www.witpress.com, ISSN 1743-3509 Similarly, though there are larger tile-to-tile variations in the Y-direction, these variations do not appear to be related to the surface appearance of the tile. Table 1. Results of dynamic modulus testing. Surface Avg. E Number Coeff. # Grooveslm of tests of Variation Tile (Mpa) X-Direction 11 l06 5.3% 14 106 8.1% 12 138 4.2% 13 165 6.5% 8 l65 7.5% Avg X dir, 12.2% 2.4.2 Compressive and flexural testing Due to limits on the available tile specimens, it was not possible to test nearly as many samples destructively, as had been conducted non-destructively. Four compression tests and four flexure tests were conducted, as summarized in Table 2. Although the number of samples does not permit true statistical comparisons, orthotropic properties were, once again, evident. On average, the X-direction compressive and flexural strengths were 34 MPa and 11 MPa, respectively, compared with 23 MPa and 5 MPa in the Y-direction. Transactions on the Built Environment vol 66, © 2003 WIT Press, www.witpress.com, ISSN 1743-3509 262 Strrcctrrral Srudir~,Rrpam atrd Marnrorancr of Hrriragr Aldzirrcruw V111 Table 2. Experimental tile strength. Surface Avg. Compr. Number of Avg. Flex. Number of Groove Strength Tile # Compr. Strength Flex tests s/m (MW tests (MPa) X-Direction l l 106 18 1 8 1 14 106 40 1 12 2 12 138 39 2 12 1 13 l65 0 11 1 8 l65 0 108 1 Avg X 34 11 dir. Y-Direction 11 l06 14 l06 12 138 13 l65 8 l65 Avg X dir.